Abstract
Rebels that fight near or capture cities gain more concessions from the government than those that remain in the periphery. Yet, not all groups challenge urban centers. Previous scholarship expects rebel strength to explain this strategic decision. However, weak rebel groups challenge cities, too. Our approach focuses on the conflict process more broadly. We argue that as the network of rebels challenging the government increases, opposition groups become more likely to attack cities as either they become emboldened, given the government’s disadvantage in multi-front wars, or they are propelled to strategic and resource centers in competition with the other groups. Statistical analysis of all African conflicts from 1989-2009 strongly supports this logic, while an exploration of most typical cases highlights each of these mechanisms in practice. This project thus links literature on civil war tactics and conflict contagion.
Introduction
Cities are crucial in civil war. They tend to be economic, education, and population hubs, to house military and police units and armories, as well as possessing symbolic value to the local population and international community. In Afghanistan, for example, U.S. counterinsurgency strategy evolved to focus primarily on protecting the country’s major urban centers while ceding portions of the periphery to the Taliban (NYT 2018). Such a move mirrors Soviet strategy in the country nearly three decades earlier when Moscow decided to protect the urban centers while indiscriminately bombing rural areas in hopes of pinning down the Mujahedeen (Dupree 1984). Rebels understand this. As the Justice and Equality Movement’s (JEM) leader Gibril Ibrahim noted, “We need to take the fight closer to Khartoum. If you fight in the desert… they don’t care” (quoted in McCutchen 2014: 20). The data support this claim. Rebels that move to the city attract more government concessions, while those that remain in the periphery receive few, if any (e.g., Greig 2015). Then, why do only some rebels move to the city?
Understanding this puzzle is critical for both policymakers and scholars of political violence, as highlighted by the focus of U.S. COIN strategy in Afghanistan and its struggles. While the capture of key locales has long undergirded our understanding of military strategy (see Smith and Stam 2004), observers are increasingly highlighting the role of urban centers in the onset of civil war (e.g., Nedal et al. 2020), its duration (Buhaug et al. 2009), terrorist tactics (Graham 2008), and civilian victimization more broadly (Koren 2017). Some are even predicting that the future of civil war will revolve around fights for mega-cities (e.g., Konaev 2019). To date, though, little research has focused on why some rebels fight near urban centers while others do not. The most straightforward answer to this puzzle is that not all rebel groups are strong enough to challenge urban areas. Only those with enough fighters and firepower, relative to the government, can advance upon these key targets (e.g., Buhaug 2010; Holtermann 2016).
But consider JEM. At its peak, the group possessed only 5000 fighters (Asal and Rethemeyer 2015) compared to the government’s 100,000+ soldiers and additional informal militia members of the Janjaweed that fought the group alongside the army (IISS 2011). Clearly, JEM was not strong relative to the Sudanese government and yet they attacked Khartoum. Therefore, while rebel strength plays some role in this calculation, we propose a broader framework to better understand these dynamics alongside the various phases of civil war and other forms of political violence. Specifically, we posit that the path of the conflict is produced by the rebels’ understanding of the costs of targeting cities—such as the government’s concentration of forces or resolve—which we label targeting costs, and the costs of remaining in the periphery—such as failing to secure important strategic locations or resources—which we label immobility costs. Such costs are not simply the inverse of each other, but are uniquely shaped by a group’s military relationship vis-à-vis both the government and other non-state groups within the broader dynamics of the conflict. Targeting costs are the costs that may come from attacking a city, while immobility costs are the costs that may come from not attacking a city.
Within this framework, we draw on a burgeoning literature noting the interconnectedness of armed group decision making within a conflict, as rebel group-government interactions rarely happen within a vacuum (e.g., Walter 2006; Stanton 2015; Uzonyi and Demir 2020). Unlike Butcher (2015) who highlights the information problem multiple groups create, we posit that rebels use each other’s violence to estimate their expected costs of moving urban. Targeting costs are those costs a rebel group must pay to attack a city. These are primarily material costs in terms of fighters, weapons, supplies, etc. that may be expended in the move towards a city and then in the ensuing fight at that location. We argue that rebels perceive lower targeting costs, and thus a greater advantage for attacking cities, as the threat to the government from other insurgent groups increases. As threat increases, the government must split its forces and resources to address each additional group. This division decreases the government’s ability to counter each group (e.g., Borman and Hammond 2016), or rest and replenish its troops (e.g., Akcinaroglu 2012), providing the insurgents more opportunity to move against valuable urban targets. We theorize on these dynamics as an Emboldening Mechanism.
However, we also note that these dynamics inform immobility costs through a Competition Mechanism in which groups are propelled to attack cities more quickly in an effort to secure valuable real estate before a rival group can control these strategic locations. Immobility costs are the costs a group may face for not moving towards or attacking a city. These capture the rebels’ strategic situation vis-à-vis other groups in the civil war. As others secure strategically important locations, and tap into the population and material resources that urban centers provide, the rebels who did not move to the city are growing weaker relative to these other groups. While attacking an urban location may not always result in victory, rebels—like all political actors—are boundedly rational. They base their calculations on the behavior of others. Seeing their rivals move urban creates fear in the group that the other rebels will capture the city and better establish themselves among the competing armed groups in the conflict—thus, the fear of increasing immobility costs pushes the group to move urban in competition with their rivals.
We test our argument through a novel quantitative analysis in which we use detailed micro-empirical analyses and apply them across a broad set of African countries, simultaneously blending work that has largely been applied by scholars to single country settings with research on large regional or global studies. We then use a series of case studies to illustrate each mechanism at work. In our statistical models, estimated across several population thresholds, we find two distinct processes at work. First, we find that an increased threat to the government posed by the presence of additional opposition groups propels the rebels in question to city centers, but only to smaller cities with fewer than 50,000 inhabitants. Second, however, if the group in question is allied to one of the other armed movements challenging the government, they become willing to assault urban centers with much larger populations. This implies that though rebels can be emboldened by observing a growing challenge to a government (Emboldening Mechanism), or make the decision to use the opportunity to secure valuable locations (Competition Mechanism), they are reluctant to target large cities without an ally. Indeed, allies provide much more reliable information about government capabilities via formal networks, and are also more apt to fight alongside the group, increasing its opportunity for war. We buttress these econometric results with cases studies that further illustrate the emboldening and competition mechanisms.
This project contributes to our understanding of political violence in two ways. First, it brings together two literature that, while focused on civil war, rarely speak to one another. We show that opportunity-based explanations for civil war (e.g., Fearon and Laitin 2003) need to account for how groups interact with each other (e.g., Maves and Braithwaite 2013; Lane 2016). Second, and relatedly, we highlight that while the opposition’s opportunity for violence undergirds its ability to challenge the government, its calculation of advantage is based on more than its own strength. These calculations are partially determined by the broader dynamic of the conflict. We believe these theoretical contributions are also important for policymakers tasked with implementing counterinsurgency policy, who may be interested in protecting cities before the fight reaches these key locations.
Rebel Advances in Multi-Front Wars
To gain concessions, rebels must inflict pain upon the government. Groups that remain unthreatening to the regime tend to receive little attention from the government. Not until a group begins to harm the government’s interests, must the regime address the rebels. Until then, they are left to toil in isolation, making little progress on their stated grievances. Since governments tend to resist making concessions to challengers, for fear of inspiring additional opposition, armed conflict tends to occur as the group becomes more active (e.g., Walter 2006). However, as the benefits of fighting begin to decrease relative to the costs of continuing the war, a government becomes more willing to make concessions to the group. The greater pain the rebels inflict, the further concessions they tend to receive as the government becomes more eager to end the war and avoid future costs (e.g., see Cunningham et al. 2009; Thomas 2014).
Rebels inflict pain upon the government by attacking military, infrastructure, and civilian targets (see Hultman 2009; Stanton 2015). Military losses hinder the government’s ability to maintain power, infrastructure destruction limits the government’s reach and hinders economic activity, and civilian victimization may turn the population against the regime that fails to protect them. Rebels seek to maximize these costs for the regime while minimizing their own losses. Thus, they look for targets that provide maximum benefit for attacking. This calculus helps explain why cities become integral to the dynamics of political violence. Cities tend to be a country’s economic and educational hubs. These large population areas often serve as the government’s tax base, and house the regime’s military and police forces (Fujita and Mori 1996; Hendrix 2010; Knight 2013). If rebels are able to target these urban locations, they can inflict high costs on the regime quickly (see Grieg 2015).
Rebels often face a problem in this endeavor, though. Such groups tend to form in the periphery of the country, among networks that support them (Staniland 2014; Carter et al. 2019). In these areas, valuable targets are limited because the rebels are far from urban environments. Thus, while some groups like al-Shabaab or the AIS and GIA form in urban centers, most rebel groups need to make their way out of the jungles or mountains to launch an urban attack. This need for movement creates two issues for the rebels. First, while the government may be content to ignore rural opposition, it cannot avoid confrontation with a group attacking its cities. Therefore, while attacking a city may inflict costs on the government, it is also risky for the rebels, entailing targeting costs for the group. For example, Milosevic was content to allow the KLA to roam the countryside, only striking at the insurgents and their civilian supporters once the group made moves against cities in Kosovo (see IICK 2000). Second, the government understands the rebels’ incentives to move urban. In response, the government tends to guard and fight over important logistical locations en route to these cities (Hammond 2018). This means that before a group can attack a city, it must first fight its way to that location.
Since rebels tend to be weaker than the government, they must be careful in moving towards a city. Following the standard logic of the loss of strength gradient (Boulding 1962), rebels may make this move when they are strong vis-à-vis the government and can fight their way to the center (Buhaug 2010; Holtermann 2016). However, a complication arises. Few groups, even with foreign support, ever reach parity with the government, let alone out power it (Cunningham et al. 2009). Yet, we see rebels emerge from the periphery to strike at the government. And, we see qualitatively weaker rebels, such as JEM, making this same push. For instance, when the NRA initially attacked Masindi in Uganda, barely half of their 7000 fighters even had weapons (Schubert 2006).
Figure 1 presents the distance to the nearest city for groups that are weaker (square points; dashed line), at parity (circle points; solid line), and stronger (square points; dotted line) relative to the government from 1989-2009 (distance measured from closest battle location). We see a general upward trend in ebbs and flows in distance as road networks and technology has improved over time. In all cases, however, we see that all groups, regardless of their relative strength, demonstrate notable amounts of movement across geographic space relative to city centers. Weaker organizations tend to display more dramatic shifts to and away from cities, likely due to group vulnerability. Though more powerful groups tend to remain closer to cities, there are moments that they shift away from city centers. Thus, we build on the existing literature to highlight high variability in groups, especially weaker groups, in terms of pushing towards urban centers. Distance to cities across differing levels of relative strength (relative to the government).
We posit that the reason for this variability in movement is twofold. First, outside of coups, rebels often lack sophisticated information on the military. They tend not to have wide-ranging intelligence gathering infrastructure, as they rely on local networks of support that may not reach into the urban center (Staniland 2014; Larson and Lewis 2018) and much rebel strategy focuses on remaining distant from military centers until fighting occurs (Walter 2009). Relatedly, rebels are unlikely to have a qualitative understanding of the military’s fighting ability—quality of the soldiers, logisticians, and field commanders. This problem is heightened when the military relies on informal militias, whose quality likely has more variance (see Forney 2015).
Second, even if rebels possess some of this information, we posit that they make their calculations about the likelihood of success based on more than the strength of their fighting force vis-à-vis the government. In part, we suggest, rebels calculate their chances of success based on the broader environment of the conflict—are there other rebel groups challenging the government and are they preforming well? As Christia (2012) underscores, rebel elites must consider a broad array of factors that may determine their daily survival. We believe rebels have a better understanding of these factors than they do of formal military strategy or strength because rebel groups often actively coordinate with one another. For example, the Eritrean People’s Liberation Front provided weapons and training to the Tigrayan People’s Liberation Front in the mid-1970’s (see Lane 2016, 396). This active coordination is not uncommon, as rebel-to-rebel assistance in terms of intelligence, logistics, territory, and weapons is one of the frequent sources of support that opposition groups receive (Högbladh et al. 2011).
However, the existence of other opposition groups provides the rebels more than a potential formal source of support. It also helps divide the government’s forces. Each new insurgent group potentially opens up an additional front on which to confront the government. Returning to the Ethiopian case, which at its peak in 1978 witnessed seven rebel groups opposing the Derg, demonstrates how multiple challengers result in multiple fronts for the government to contain. There, for example, the WSLF attacked from the east, the EPRP from the center, the OLF from the east and west, the ALF from the north-east, the TPLF from the north-central, the EDU from the north-west, and the EPLF from the north. As Akcinaroglu (2012) highlights, these multiple fronts put the government at a disadvantage vis-à-vis the collection of rebels, even if the groups are not actively or formally coordinating. First, multiple fronts force the government to divide troops and resources rather than concentrate on eliminating any one group. Second, battles tend to occur temporally close together such that the government cannot rest or regroup while each rebel group can while waiting for their time to fight. Lastly, the fronts are often geographically diverse such that troops are not trained to move across fronts, limiting the government’s ability to replenish divisions that take large losses. As the number of rebel groups opposing the government increases, this shifts the probability of victory for the rebels in each given location of the country.
We believe this logic creates the expectation that as the number of rebel groups actively challenging a government increases, especially if rebels are formally coordinating, so does the likelihood that any one of them attacks a city in a given period. However, we also posit that two complementary mechanisms drive this logic. We term Mechanism 1 the Emboldening Mechanism. Here, the various groups’ anti-government activity provides both formal (when actively coordinating) and informal (when not actively coordinating) information about reductions in the government’s strength. Observing the government’s growing disadvantage from confronting several groups at once, each given group is emboldened to believe their costs to targeting a city are decreasing.
Mechanism 2, which we call the Competition Mechanism is also at work. Here, rebel groups that are in active opposition to each other rush to city-centers to beat their enemies to valuable resources or key strategic locations to survive the day (see Christia 2012). In this situation, as with Mechanism 1, the groups are informally learning about their opportunity vis-à-vis the government from their competition. 1 But, their move against a city is propelled by a fear of losing an opportunity to take the valuable location before the other opposition groups. Thus, a group’s immobility costs are increasing alongside any additional non-state actors in the conflict.
We note that it is not always better for a rebel group to beat another group to an urban location—arriving first does not mean that the group will win or pay lower targeting costs. However, each rebel group is boundedly rational and faces an information problem regarding its probability of victor and likely costs for each potential battle. In part, a rebel group tries to solve this information problem by watching the movements of the other groups. A rival often interprets its enemy’s moves as more threatening than those of another actor (e.g., Uzonyi and Rider 2017), and becomes paranoid that its rival’s move will decrease its military and strategic position relative to that enemy (e.g., Thompson 2001). For this reason, the group will react more forcibly to the actions of its rival (e.g., Lektzian et al. 2010). It is for this reason that we posit that the rebels will race their rival to the urban center—when the group sees its rival moving, it comes to believe that their targeting costs are lower than they originally expected and that their immobility costs are now increasing, given their rival’s movements.
Together, the emboldening and competition mechanisms provide our hypothesis.
An opposition group moves closer to a city as the threat posed to a government by other rebels increases during the conflict.
Research design
To evaluate our theory, we build a rebel-month dataset encompassing groups active in Africa from 1989-2009. The temporal range is determined by our key variables of interest. Rebel groups are defined using the approach advocated by the UCDP/PRIO Armed Conflict Dataset (Gledistch et al. 2002; Pettersson and Oberg 2020). To be deemed a rebel organization, the group in question must have an announced name, be organized, and capable of waging effective resistance against the state. Further, to be “active” the group must be actively involved in a “contested incompatibility” with the state that has produced at least 25 battle-related deaths in a calendar year. In total, our sample includes 81 rebel groups operating in 28 different African countries. 2 Rebel group campaigns against the state averaged 22.93 months, with a low of 13 and a high of 186. 4 In total, our sample consists of 1857 rebel-months.
Our dependent variable is
The latter piece of necessary information, the location of African cities, is derived from a freely available database created using census results and official estimates.
7
No set definition of a city exists. For our purposes, we estimate models using a sample of all African cities, and a series of thresholds to assess how population size influences being targeted by rebel forces. The cities included are visualized in Figure 2. The size of the spatial point corresponds to the number of inhabitants, while capitals are designated using triangle symbols. As noted, we use the location of these cities to measure the distance from armed non-state actors to urban centers. Measured in kilometers, the natural log is taken to correct for skew. The nearest city variable (when not imposing a threshold) ranges from 0 – 5.782, has a mean value of 2.719, and a standard deviation of 1.730. The capital city distance variable ranges from 0 – 7.014, has a mean value of 2.042, and a standard deviation of 1.534. Cities across Africa. Size of point corresponds to population, triangle indicates capital city.
Our theory expects that a rebel group moves closer to a city when other opposition groups are threatening the government—thus distracting the government and forcing it to divide its forces. We capture this threat by measuring the intensity of government battles with other rebel groups fighting in the country. We construct this measure in two steps. First, for a given rebel group, we identify the number of other, non-allied groups active in the same country in a given month. We label this set of other groups as the rebels’ “network” and exclude any allied rebel organization. Second, for each network, we count the number of total battle deaths that occur as a result of government-rebel battles and take the natural log to correct for skew. Because we anticipate the evolution of events over several months to influence behavior, we use the 3-months moving average of battle deaths in the network as our key independent variable,
Our second independent variable of interest aims to account for the formal process (coordination) of the Emboldening Mechanism. We anticipate rebel alliances to influence the willingness of groups to move to city centers. To operationalize
We also control for seven factors that we expect to explain variation in the change in the distance of active rebel groups to city centers from the prior period: whether or not the target is a
We also introduce country fixed-effects by including a set of dummy group variables each multiplied by its regression coefficient. Country-fixed effects are a useful tool because they not only capture unobserved time-invariant factors driving group behavior, but also because they control for time-invariant factors that may be important in explaining distance to city centers, such as the geographic size of the country, the density of city centers, terrain, transportation networks, among others. This also helps account for measures that are captured annually and are thus constant in much of our monthly research design.
An additional concern is unexpected variation occurring due to events transpiring during a given temporal period. To prevent this from driving our findings, we also introduce year and month dummy variables into our equation that are multiplied by its regression coefficient (time-fixed effects). 19
In sum, our model is specified as follows, where distance of group i to city j at time t is a function of our independent variables of interest,
Findings
Summary of time fixed-effects regression models across population thresholds.
Our key variable of interest is Other Rebel Violence. This variable captures the informal process described in the Emboldening Mechanism and the logic put forth in the Competition Mechanism. Looking at the relationship between Other Rebel Violence and Distance to Cities, across population thresholds in the definition of “city”, we find an interesting pattern: the size of the city matters greatly. The higher the number of battle deaths within a group’s network, the closer they are located to a city center when the population is relatively small (≤ 100,000). As the population increases, however, we find no relationship between the intensity of government battles with other groups and where rebels in question are located relative to city centers. This reveals that though increased violent interactions between the government and other rebel groups do lead to a decreased distance to city centers, this applies to smaller cities and not larger urban centers. This may be driven by the fact that larger cities are characterized by dense transportation networks and a larger government presence, substantially raising the potential costs for rebels. There is also a need to hold a city after capture, which may be difficult given an aggressive campaign by a government to recapture a large, and thus valuable city. In this sense, rebels are apt to balance the benefit gained from approaching a city to the potential backlash following capture and subsequent occupation.
The Rebel Alliance variable, capturing the formal component of the Emboldening Mechanism whereby fatalities emerging from battles involving an allied group leads the organization in question to move towards city centers to fight alongside their ally (or, target other cities to further pressure the state), finds mixed support. The variable fails to reach statistical significance (p ≤ 0.05) across most population thresholds, except for 20,000, 25,000, and 300,000 inhabitants. In smaller cities, we see increased violent government interactions with allied groups to push rebels away from city centers, but the relationship shifts when larger cities are considered. There is limited evidence, therefore, that larger urban areas are more apt to be targeted when an allied group is confronting the government. Though, given that this relationship is not statistically significant in most population thresholds, the evidence for this finding is limited.
Among the control variables, Distance to City During the Prior Period, Rebel Strength and Capital City are the most consistent: more distant rebel organizations are more likely to remain so, stronger rebel organizations are expected to be located closer to city centers, and capital cities are more apt to attract armed movements. These patterns have been confirmed by prior research. Importantly, including rebel strength indicates that the rebels’ network captures more than a simple aggregation of rebel strength vis-à-vis the government. As expected, our models also reveal that across a majority of population thresholds rebel groups with Secessionist goals are less apt to target cities, as are insurgents operating in a country where a UN Peacekeeping mission is deployed. Interestingly, both Government External and Rebel External Support appear to shift directions based on the population of the targeted city, with the former leading rebels to target smaller cities, and the latter drawing rebels to larger cities. Finally, rebel organizations with a Political Wing appear more likely to be located near larger urban areas, but there is no relationship when investigating the spatial location of rebels to smaller city centers.
While the general direction of these relationships provides valuable insight, it is necessary to determine the substantive effects. To accomplish this, first we plot standardized beta coefficients to assess both the strength of the effect, and how it changes across population thresholds. These are found in Figure 3, with triangle points indicating Other Rebel Violence and circle points the presence of a Rebel Alliance. We also include the effect of Rebel Strength as a baseline for comparison, because it is the strongest predictor in our models and has been established as the most important factor in prior research (square points). The corresponding Gay bars represent 95% confidence intervals for the estimates. First, looking at Other Rebel Violence, we find a strong effect at lower population thresholds, and a shift from a negative to a positive relationship as the population of a city increases. This, again, reveals that rebels become willing to challenge smaller cities while the government is engaging other rebel movements, but groups remain reluctant to assault larger urban centers. The effect of this variable is noteworthy, being equal in statistical terms to Rebel Strength (relative to the national government) at each of the population thresholds where the variable is statistically significant. We conclude, therefore, that the presence of other challengers, especially when they are actively fighting the government, is important in understanding why rebel movements shift their focus to city centers.
22
Relative Effects of Covariates Included in Regression Analyses (standardized beta coefficients).
Probing the Mechanisms
Our large-N statistical analysis provided strong support for our argument. Rebels tend to move towards cities when other rebel groups are simultaneously confronting the government, and towards larger urban centers when an allied group is actively challenging government forces. While we can capture the number of groups fighting empirically, our regressions are unable to measure directly the two mechanisms that we propose drive this strategic positioning (Emboldening and Competition). In addition, our models do not capture which specific cities rebels are moving towards. This has important implications, and therefore, warrants further investigation.
To begin, as seen in Figure 4, we plot the distance to the nearest city (ln) and the nearest networked rebel organization (ln) against one another. As such, only rebel-months in which at least two rebel organizations were challenging the same government appear in the plot. In addition, capital cities are assigned square symbols, while all other cities are circles. The size of each point represents the total population of the city and labeled groups are those who are allied with another group that is fighting during the same period. Rebel Distance from Networked Group and Nearest City (rebel-months). Allied groups are labeled. Point size indicates population of city.
This leads to four quadrants: (1) two or more rebel groups fighting near the same city (lower-left), (2) a rebel group fighting near a city but far-removed from other rebels (lower-right), (3) two or more rebel groups located near one another and not located near cities (upper-left), and (4) rebels who are distant from both cities and other rebel organizations (upper-right). The dominant position among rebels operating in the same country as other organizations is the lower-right quadrant, indicating that during the majority of rebel-months rebels prefer to locate near city centers but avoid other groups. We see a clustering of rebel alliances in the bottom portion of this quadrant, indicating having an ally makes it more likely that the group is located closer to a city, even if its ally is not nearby. There are several large cities, including capital cities, in this quadrant, as well.
The next most common quadrant is the upper-right, which are rebel groups existing far removed from both cities and other groups. This is consistent with the argument in the literature that rebels often form and exist in the periphery, waiting to accumulate power prior to making their move towards the “center.” It is also worth noting that closest city to groups in this quadrant are comparatively small in terms of population, again hinting that they are strategically avoiding areas of high state-capacity. Though we cannot confirm using our data, it is likely these are locations characterized by rough terrain and dense forests, as these areas are far-removed from cities and provide a sense of physical security from a state’s military forces. Interestingly, there is a large cluster of allied groups: behavior that may be incentivized due to relative power differences.
The lower-left quadrant represents rebels that have moved to the “center.” That is, they are near at least one other rebel movement and a city. The proportion rebel alliances in this quadrant are the highest (excluding the upper-left, which is rare positioning). This is not unexpected given the risks associated with operating near larger cities. We find all these alliances, however, when rebel groups are located near city centers. This observation, along with the cluster of alliances near the bottom of the lower-right quadrant, highlight one of the core findings from our regression models: rebels are apt to fight near city centers, especially near larger urban areas, when they are in an alliance. This figure reveals that while allies often target the same city, they are also more willing to move towards other larger cities away from their ally.
The final quadrant, found in the upper-left, are groups located in close proximity but far-removed from city centers. This is, by far, the rarest of all positioning. As this is a sample of armed non-state groups who are in a context where a government is being challenged by at least one other rebel group, this is not to say that this positioning is not observed often in dyadic conflicts (government vs. single rebel group). Rather, when multiple potential challengers are present, this is an unlikely strategy. Rather, rebels who prefer to reside in the periphery do so in isolation (upper-right) and not alongside other groups.
This brief analysis provides additional insight into whether rebels are targeting the same city, or choosing to move towards different urban areas while the government is confronting an increasing threat via another armed non-state actor. While the lower-right quadrant may imply that the Emboldening Mechanism is at work because rebels are choosing to target different cities while government forces are diverted fighting another group, this may also be the case for the lower-left quadrant. Indeed, rebels may choose to lash out at government forces while they are struggling to cope with another armed non-state actor in the same location. Or, as we see in Figure 4, a group may choose to fight alongside their ally. The same issue arises when attempting to differentiate the Competition Mechanism: rebels may race to a city in an attempt to capture strategic locations, and these locations may be where other groups are located (lower-left quadrant) or removed from other movements (lower-right quadrant).
To further verify the validity of our claims, we follow Seawright and Gerring (2008, 297) and use the “most typical” criteria to explore a series of brief illustrative examples to help verify the validity of our claims. In particular, we use each case to trace where there is evidence of either of our two mechanisms—Mechanism 1 (Emboldening): rebels have an offensive advantage as they divide the government’s forces and bog them down in frequent fighting, and Mechanism 2 (Competition): rebels are drawn to the city in a race for resources or strategic locations.
Ogaden National Liberation Front
The Ogaden National Liberation Front (ONLF) represents ethnic Somalis living in the border region between Ethiopia and Somalia, and has engaged in a violent irredentist claim to unite Ethiopian-Somalis with Somalia since the 1990s. However, the group produced little success for over a decade, as its small size (∼2000 fighters) prevented it from making a push out of the periphery of Ethiopia’s Somali region. However, in 2007, the ONLF’s fortunes changed as it launched a series of attacks on increasingly large towns, culminating with an attack on the city of Jijiga (HRW 2008). The African Research Bulletin (ARB 2007) links the ONLF’s move out of the periphery directly to developments in the interlocking cross-border civil conflicts of Ethiopia, Eritrea, and Somalia. With the ARS/UIC, or Islamic Courts Union, escalating its attacks against Mogadishu in Somalia, Ethiopia sent troops to defeat the rebels. “When the Ethiopian National Defence Force (ENDF) went into Somalia against the Islamic Courts Union in late December 2006 and then occupied the capital, Mogadishu, it used many of its units from the Ogaden. The ONLF quickly exploited their absence, launching increasingly bold attacks on military barracks and towns, culminating in the Abole oil site attack” (ARB 2007, 17,240). The description provided by ARB supports our claims and captures Mechanism 1 well. Multiple fronts forced Ethiopia to divide its troops, pulling them out of Ogaden, and providing opportunity for the ONLF to press forward.
Justice and Equality Movement advances on N’Djamena and Khartoum
In 2000, the Justice and Equality Movement (JEM) formed to overthrow the dictatorship of Omar al-Bashir and correct structural inequalities in the country. By 2003, the group was engaged in warfare throughout the deserts of Sudan’s Darfur Region. Due to the group’s small size (∼5000 fighters) and rural formation, JEM remained in the periphery. However, in 2008, the group changed strategies. First, in February, JEM moved across the border to fight in N’Djamena. Chad’s President Déby had long supported the rebels against his rival Sudan, and was a reliable source of valuable resources for JEM. Now, Déby was under attack by Sudanese troops and several of their rebel protégés. JEM raced out of the desert to protect N’Djamena and engaged in its first taste of urban combat against this threat to its resource well (Reuters 2008). JEM’s advance on N’Djamena to protect this strategic location and resources from its opponents offers support for Mechanism 2. Here, JEM’s rivalry with a network of Chadian rebels—the UFDD, UFDD-F, and RFC—propelled it into urban conflict. El-Tom (2013, 73) reports that while JEM’s leadership had long discussed the possibility of urban fighting, the developments in N’Djamena brought it into an urban environment earlier than it had initially planned.
Rebel Coordination in Darfur
After fighting in N’Djamena, JEM moved to attack the capital of Sudan, Khartoum. However, they were turned away in the capital suburbs of Omdurman. Following this loss, JEM retreated into the desert of Darfur and remained in the periphery. However, JEM’s commanders had been close to Khartoum and still believed that moving to the capital was the best chance the opposition had to remove al-Bashir. The problem was that the group was too small and the government could concentrate its forces on any urban attack. In response to this dilemma, JEM began to hold summits throughout Darfur, attempting to ally itself with other rebel groups and factions. Ibrahim’s goal was to coordinate rebel efforts to open multiple urban fronts and fight battles temporally close together, to exhaust the government’s forces. His vision became popular among the Darfurian rebels and JEM eventually convinced the region’s other primary rebel group, the Sudan Liberation Movement/Army (SLM/A), to conduct joint operations against al-Bashir’s regime (see ARB 2008, 17,690). Prior to this coordination, SLM/A had remained in the periphery due to its inferior strength, as well. However, once coordinating with JEM, the group moved quickly to attack the city of Kutum in 2008. JEM followed these attacks with its own advances on Muhajiriya and Gereida a few months later (Sundberg and Melander 2013). This case supports our hypothesis and illustrates Mechanism 1 well: rebels coordinate and share information to divide the government’s forces and keep them frequently engaged in warfare.
Union of Democratic Forces for Unity
In the cases thus far, the rebels were often isolated in the periphery until other groups formed or entered the sphere of fighting. However, the move to the city can happen quickly, as well. Consider the Union of Democratic Forces for Unity (UFDR). In 2006, the UFDR formed amongst the Goula ethnic group in the Central African Republic’s (CAR) northeastern city of Birao, far from the capital of Bangui. The rebels claimed that President Bozize governed the country to favor his Baya ethnic group at the expense of all others and aimed to force change (ACLED 2009). With the government occupied with fighting against the Popular Army for the Restoration of the Republic and Democracy (APRD) in the northwest along the Chadian border, the UFDR was able to move south, uninhibited, through CAR’s diamond producing area. There, the group gathered wealth and additional fighters before marching on Bria, the capital of the Haute-Kotto Prefecture (ICG 2010). In the SLM/A case, the rebels actively coordinated to divide the Sudanese government’s forces. In CAR, the rebels were not active collaborators. Instead, Mechanism 1 worked through the informal channel. The UFDR’s move from Birao to Bria happened within months, thanks to the APRD bogging down CAR’s forces in the west at the time of the UFDR’s formation.
United Front for Democratic Change
The Rally for Democracy and Liberty (RDL) is a Chadian rebel group that combines both the formal and informal logic of the emboldening mechanism (Mechanism 1). Amidst the cross-border chaos of the interlocking CAR, Chadian, and Darfurian conflicts, the Sudanese government formed the RDL in the end of October 2005 and ordered it to attack Adré in Chad. Chadian troops easily repelled the nascent RDL back into Darfur. From there, though, the RDL took actions that made use of both the active coordination and passive observation portions of Mechanism 1. Mahamet Nour Abdelkerim, the RDL’s commander, began formally coordinating with several small Darfurian groups under the banner of the United Front for Democratic Change (FUC). He then waited until fighting escalated once more on Chad’s eastern border between other groups before cutting through CAR to move the RDL/FUC urban and attack N’Djamena. As Massey and May (2006: 445) recount, “Taking advantage of the Chadian army’s vulnerability with the majority of its troops massed on the eastern border… the FUC launched an offensive to N’Djaména.”
Discussion and Conclusion
Why do only some rebels move to the city, when attacking such valuable locations increases the likelihood that the government will grant them concessions? We have shown that—regardless of their own strength relative to the government—rebels tend to advance towards smaller cities when the government is embroiled in conflict with other groups, and will assault larger urban areas especially if actively coordinating with an ally. Such multi-front wars require the government to divide its troops, providing an opening for rebels to target city centers. While these findings are consistent with our theoretical expectations, and buttressed by several illustrative examples and a deeper exploration into our data, several questions remain. Here, we focus on three that speak to limitations in our research.
First, where are battle taking place? In this study, we do not account for where the violence occurs, even though we did explore this issue in a limited way. As the number of rebel groups fighting the government increases, rebels may either move to the area where this fighting is occurring in an attempt to “tip the scales,” as it were. Or, they may decide to attack cities far removed from the conflict zone. This decision is likely predicated on the nature of relationships among active rebel organizations: the presence of a non-state rivalry incentivizes an attack removed from the conflict zone, whereas cooperative relations lead to an attempt to “tip the scales” against the government at a particular location. Accounting for where violence is taking place, rather than solely focusing on when, will allow future scholars to better understand how the relations among non-state groups influences their war making strategies. The literature clearly notes that in multi-front wars rebels often fight one another more often than the government (Cunningham, Gledistch, and Salehyan. 2009, 572). Rebels compete for resources and political capital to aid their fight against the government (Fjelde and Nilsson 2012; Hazen 2013). This behavior has been observed in Afghanistan, Liberia, the Democratic Republic of the Congo, Burma, and Sudan. These non-state interactions therefore, are vital to investigate moving forward.
Second, which cities do rebels attack? Not all cities are equal. When rebels perceive an opening that allows a move to a city center, they must first decide which city to target. This decision is predicated on the strategic importance of the city, but some targets may prove costlier than others may, as evident by how our results change across city population thresholds. Scholarship has established that some factors, such as rough terrain, forests, international borders, and precipitation that undermines transportation networks can provide rebels with physical security and a means of escape if they become overwhelmed. Because of this, a majority of rebel groups prefer security-enhancing features (Reeder 2018). This allows rebels that are usually weaker relative to the national government and thus vulnerable to protect the long-term survival of the organization. Cities that are characterized by security-enhancing features, then, may be more likely to be targeted by rebels while other groups are fighting, especially when the group in question is much weaker relative to the government. Further effort is needed to more fully unpack this decision calculus and better understand the differences that emerge across population thresholds, and other important factors.
Lastly, how might the decision to move urban in dyadic wars differ from the multi-party context on which we focus? Previous research tended to focus on the decision calculus of singular rebel groups in dyadic wars against the government (e.g., Holterman 2016 on CPN-M vs. Nepal). In this environment, the rebels are unable to ally with other groups or move urban based on broader trends in the conflict. However, we expect that the mobility and immobility costs for such groups are still likely influenced by larger contexts. Coups, interstate war, or the promise of foreign support may each alter the rebels’ strategies at various points in the war (for example, see Leventoglu and Metternich 2018). We urge future scholars to better integrate our models of rebel decision making across these various contexts to provide a complete understanding of the process of political violence.
Supplemental Material
Supplemental Material - Capture the Fort: Explaining the Timing of Rebel Assaults on Cities During Wartime
Supplemental Material for Capture the Fort: Explaining the Timing of Rebel Assaults on Cities During Wartime by Gary Uzonyi, Bryce W. Reeder in Journal of Conflict Resolution
Supplemental Material
Supplemental Material - Capture the Fort: Explaining the Timing of Rebel Assaults on Cities During Wartime
Supplemental Material for Capture the Fort: Explaining the Timing of Rebel Assaults on Cities During Wartime by Gary Uzonyi, Bryce W. Reeder in Journal of Conflict Resolution
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Supplemental Material
Supplemental material for this article is available online.
Notes
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
